13. Mixed Methods as Methodological Integration

Mixed methods is a design decision

A study that includes a survey and interviews is not automatically a mixed-methods study. Mixed methods is a design decision — the integration of qualitative and quantitative strands is constitutive of the question. Without integration, you have a survey-plus-interviews study, which is fine, but not the same thing.

Three tests for whether your study is mixed:

1. Could you delete one strand and still reach the same conclusions? (If yes, not integrated.)
2. Do the strands inform each other in design, sampling, or analysis? (If only in interpretation, integration is thin.)
3. Is there a planned integration product — a joint display, a meta-inference, a side-by-side comparison structured for synthesis?

If the answers are no/no/no, the study is multi-method, not mixed.

Pragmatism as the philosophical anchor

Mixed methods sits philosophically uncomfortable for paradigm purists. Positivism and constructivism disagree on what counts as evidence; mixing them seems to require choosing one.

Pragmatism resolves the discomfort by changing the question. Rather than ask "which paradigm is right?" pragmatism asks "what does the question demand?" The criterion for method choice is fit to question, not paradigm orthodoxy.

Pragmatism is the dominant philosophical anchor for mixed methods, though not the only one. Critical realism, transformative paradigms, and dialectical pluralism are alternatives. The point is to name your anchor explicitly. An unstated philosophical stance is still operative, just hidden.

The major mixed-methods designs

Three core designs, each with a distinct purpose:

Convergent design

Both strands run in parallel. Results are compared, contrasted, and integrated to triangulate or surface complementarity.

Purpose: when you want a single integrated picture at one moment in time, using both methods to illuminate different facets of the same phenomenon.

Example: a survey of caregiver well-being plus interviews on the same population during the same period. Joint display compares quantitative scores with qualitative themes.

Explanatory sequential design

Quant first, then qual to explain the quant findings. The qual sample is purposively drawn based on quant results (e.g., extreme scorers, outliers, surprising patterns).

Purpose: when you have a quantitative finding that needs explanation — heterogeneity, unexpected pattern, mechanism unclear.

Example: a randomized trial shows reduced anxiety but variable effects. Follow-up interviews with high responders and non-responders to explain why.

Exploratory sequential design

Qual first, then quant to test or measure. The qual phase informs instrument design, hypothesis generation, or measurement development.

Purpose: when no adequate instrument exists, or when the phenomenon needs framing before measurement.

Example: ethnographic study of a community's experience of a chronic condition, used to develop a culturally validated outcome measure, followed by a survey using the new instrument.

Choosing among them

The purpose drives the choice:

• Triangulation at one moment → convergent
• Explanation of a quant finding → explanatory sequential
• Instrument development or framing → exploratory sequential

Sometimes a study needs an embedded design (one strand nested in the other) or a multiphase design (sequential cycles over a long project). These are extensions of the three core types.

Integration points

Integration can happen at four points:

• Design integration — strands inform each other's design (qual themes shape survey items; survey results shape interview sampling)
• Sampling integration — purposeful sampling links strands (qual sample drawn from quant extremes)
• Analysis integration — joint displays, meta-inferences, transformed-data analysis
• Interpretation integration — final integrated narrative

A well-designed mixed study plans integration at multiple points, not only at interpretation. The integration product is specified in advance.

Joint displays as integration tool

A joint display is a table or figure that puts quant and qual findings side by side, organized by a shared dimension. Examples:

• Rows = subgroups; columns = quant indicator, qual theme, integrated inference
• Rows = themes; columns = qual quote, quant prevalence, integrated meaning
• Rows = time points; columns = quant trend, qual narrative, integrated trajectory

The discipline of building a joint display forces synthesis claims. If you can't fill the "integrated inference" column, you don't yet have integration — you have parallel findings.

Meta-inferences

A meta-inference is a claim that neither strand could make alone. It integrates the qual and quant evidence into a higher-order statement about the phenomenon.

A meta-inference should be:

• Stated as a falsifiable claim
• Supported by both strands' evidence
• Distinct from either strand's stand-alone conclusion
• Useful for the audiences the study is for

Example. Quant: peer-led education reduced vaping uptake by 15% in suburban schools. Qual: in identity-driven settings, vaping was performative; education didn't reach the performative motivation. Meta-inference: educational interventions show promise in informational-deficit contexts but are unlikely to scale uniformly into identity-driven contexts; context-specific design is required.

That meta-inference does work neither strand could do alone.

A worked vignette

A team studies adherence to a long-acting injectable medication for severe mental illness. They use explanatory sequential design.

Quant phase: cohort of 200 patients tracked for 18 months. Adherence drops sharply around month 9 for a third of the cohort.

Qual phase: in-depth interviews with 12 month-9 dropouts and 8 sustained-adherence patients. Theme: dropouts experienced a "stability illusion" — feeling well led them to attribute wellness to themselves rather than the medication, prompting discontinuation.

Joint display: rows = patient profile (sustained, early dropout, month-9 dropout); columns = quant trajectory, qualitative attribution pattern, integrated implication.

Meta-inference: a planned attribution check at month 6 — a clinician-led conversation about who or what is keeping the patient stable — may reduce month-9 dropout by addressing the stability illusion before it triggers discontinuation. The meta-inference generates the next study and a practice change.

Closing

Mixed methods integrate; co-occurrence is not enough. Pragmatism is the typical anchor — name yours. Convergent, explanatory sequential, and exploratory sequential designs each have purposes. Integration happens at design, sampling, analysis, and interpretation. Joint displays force synthesis; meta-inferences make claims neither strand can make alone.

Next: advanced integration and joint displays — process integration across the lifecycle and managing paradigmatic tension productively.